In this work, we present an additive model for space-time data that splits the data into a temporally correlated component and a spatially correlated component.

In this work, we present an additive model for space-time data that splits the data into a temporally correlated component and a spatially correlated component.

This paper studied the graphical structure in spatiotemporal data through transferring the time series data into an additive model, by assuming stationary temporal correlation structure and time-varying undirected Gaussian graphical model for spatial correlation structure. With the assumption that the spatial correlations change smoothly with time, they proposed estimators for both spatial and temporal structures based on kernel method and GLasso approach. The statistical convergence property of the estimators was provided under certain assumptions. The approach presented good performance in both simulation and fMRI data application studies. This paper is overall clearly written, with solid theoretical support and interesting application.

In this work, we present an additive model for space-time data that splits the data into a temporally correlated component and a spatially correlated component.

In this work, we present an additive model for space-time data that splits the data into a temporally correlated component and a spatially correlated component. Under assumptions on the smoothness of changes in covariance matrices, we derive strong single sample convergence results, confirming our ability to estimate meaningful graphical structures as they evolve over time. We apply our methodology to the discovery of time-varying spatial structures in human brain fMRI signals. Papers published at the Neural Information Processing Systems Conference.

In this work, we present an additive model for space-time data that splits the data into a temporally correlated component and a spatially correlated component. We model the spatially correlated portion using a time-varying Gaussian graphical model. Under assumptions on the smoothness of changes in covariance matrices, we derive strong single sample convergence results, confirming our ability to estimate meaningful graphical structures as they evolve over time. We apply our methodology to the discovery of time-varying spatial structures in human brain fMRI signals.